Gemini Omni 1.2 Flash Launches: 15-Second Videos, Lower Costs, and Better Consistency

Gemini Omni 1.2 Flash marks a significant advancement in AI video generation, offering professional-grade capabilities that are both highly detailed and significantly more cost-effective for large-scale content production. This new iteration is rolling out across the Gemini API, Google Vids, and Gemini for business, making advanced video creation accessible to a broader range of users and industries.

At its core, Omni 1.2 Flash introduces native support for generating videos up to fifteen seconds in length, a substantial increase from the ten-second limit previously offered by Omni 1.1. This extended duration is crucial for professional workflows, allowing creators to tell complete stories, develop detailed product explainers, or produce full social media advertisement spots within a single, continuous generation. Furthermore, users now have granular control over the video length, specifying durations in one-second increments from three to fifteen seconds, providing the necessary flexibility to structure complex narratives.

One of the most transformative features is the implementation of multi-turn conversational editing. Previously, making adjustments often required starting the entire generation process from scratch, which was time-consuming and resource-intensive. With this new capability, users can now refine clips through follow-up conversations. This means that if a creator needs to swap out a product, change the background environment, adjust the lighting conditions, or modify the aspect ratio, they can do so turn by turn. This drastically streamlines the production pipeline, enabling the rapid creation of multiple ad variants for different audiences and placements without the need for extensive re-briefing or re-generating the entire sequence.

Beyond the structural improvements, Omni 1.2 Flash boasts vastly enhanced reference adherence and consistency. This is particularly critical for commercial applications, such as virtual try-ons or branded advertising. The model can now guide generation by combining multiple inputs—including character images, product shots, style images, short motion clips (up to three seconds), and specific audio cues—and maintain these details flawlessly from the first frame to the last. For instance, when generating a 15-second virtual try-on sequence across 360-degree turns, the model can keep the model's identity, the specific cut and drape of each garment, and even the back details consistent, even when combining front and back product shots. This level of sustained consistency minimizes visual 'drift' or 'morphing' across camera movements and transitions, which was a common challenge in previous AI video models.

The model's commitment to consistency is further demonstrated in complex scene building. For example, when assembling a poster design, Omni 1.2 Flash can build the scene sequentially, layer by layer, ensuring that every material, shadow, and typographic detail remains locked and consistent through the final frame, even when dealing with intricate elements like faceted crystal or coiled-cord letterforms. This reliability is essential for high-stakes commercial and artistic projects.

From a technical and economic standpoint, the release is highly beneficial. Google has reduced the price by 30% compared to Gemini Omni 1.1, making the service available at $0.07 per second for 720p video. This cost reduction, combined with an optimized workflow, allows professionals to produce high-quality video content at a fraction of the previous cost. The workflow itself is designed for efficiency: users can draft initial concepts quickly at 360p or 720p, make precise, localized fixes using multi-turn editing, and only then upscale the best resulting clips to high resolutions like 1080p or 4K. This tiered approach ensures that computational resources are used efficiently, focusing high-resolution processing only on the final, polished output.

Furthermore, the model enhances audio and character performance capabilities. Users can now generate scenes with more fluid, natural-sounding dialogue and clearer lip-syncing. The system supports distinct multi-speaker conditioning, allowing it to route sampled voices to the correct characters with natural comedic timing across an entire 15-second animated scene. This capability is vital for creating polished, professional explainers or multi-character narrative videos. The ability to direct complex, continuous camera movements, such as sweeping FPV flythroughs or dramatic zoom-outs, is also significantly improved, resulting in fewer unwanted jump cuts, steadier scale, and a more polished, cinematic viewing experience overall. Overall, Gemini Omni 1.2 Flash represents a paradigm shift, transforming AI video generation from a novelty tool into a reliable, scalable, and economically viable production asset for professional creative industries.